Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
NOVA ScienceNOW

Art Authentication: Expert Q&A

  • Posted 07.08.08
  • NOVA scienceNOW

On July 8, 2008, Eric Postma answered questions about how computer analysis can help detect the non-genuine article from a group of otherwise legitimate paintings and other aspects of art authentication. 

Eric Postma

Eric Postma

Eric Postma is a Professor of Artificial Intelligence at Maastricht University in the Netherlands. Full Bio

Photo credit: Courtesy Eric Postma

Eric Postma

Eric Postma is a Professor of Artificial Intelligence at Maastricht University in the Netherlands. His research focuses on human perception and cognition, and how artificial intelligence (AI) can model these processes. In his applied research, he collaborates with the Van Gogh museum in Amsterdam and an international consortium of researchers on the digital analysis of paintings to support art historians in their studies. On the personal side, Postma reports that he has two teenaged sons and, like many Dutch, rides a bike (without wooden shoes). Unlike most of his countrymen, he doesn't like soccer. Science is both his vocation and his hobby.

Q: This episode showed the computer detecting the hesitation and overpainting arising when someone tried to exactly copy an existing painting. How well do your algorithms do when an expert forger, someone who has already made a study of the artist's style, paints an entirely new, "undiscovered" painting? Pat Sitton, Corvallis, Oregon

Eric Postma: Dear Pat,
The best example of such a case is the Wacker forgery (The Sea at Saintes Maries), which also appeared in the show. This forgery, one of a series of forgeries by Wacker, was detected by our algorithms in the same way that the algorithm was able to detect the deliberate copy that Charlotte Caspers made for NOVA. The Wacker forgery also had too many prominent brushstrokes. Of course, it would be interesting to know if any forged paintings are currently believed to be genuine. In case our algorithms detect "strange patterns" (too many brushstrokes is only one of such strange patterns), we certainly will consult the experts of the Van Gogh museum and ask them to have another look at the painting. We hope to be able to make such suggestions soon to help discover a new genuine or fake Van Gogh.

Q: Are museums eager to use computer technology to uncover fakes they might own, or do they shy away and adhere to "don't ask, don't tell" in order to keep their collections complete? San D. Hasselman, Berkeley Heights, New Jersey

Postma: Dear San,
Most (older) art historians are not familiar with the strengths and limitations of computer software. This certainly slows down the acceptance of our software in the domain of art history. Having said this, our experience with the Van Gogh Museum and the Kröller Müller Museum has been very positive, and we have also had good discussions with curators at the Museum of Modern Art (MoMA) in New York. The experts at these museums are very open-minded and sincerely interested in the new technology. Their only (fully justifiable) hesitation is based on the fact that our algorithms are still under development. We are currently extending and evaluating our algorithms in close collaboration with art historians, and we hope to be able to train new generations of art historians on the use of digital techniques.

Q: Dear Eric,
I have had in my possession for 15 years a beautiful painting signed Winslow. After doing much research on the internet, I find that there are no known artists with the sur-name of Winslow, who ever painted a seascape, which this painting is. The only other alternative is the master American painter Winslow Homer. However, all of the so-called Winslow Homer experts that I have contacted think that it is "silly" that Winslow Homer would ever sign a painting by simply signing it "Winslow," and they discredit the painting with out any further inspection or consideration. Could your process ever solve this mystery? Sincerely, Andrew Papa Andrew Papa, Chicago, Illinois

Q: Hello Professor Postma,
What a fascinating topic! Are you able to analyze an oil painting assumed to be painted by Joseph Mallord William Turner? The painting has been in our family for more than half a century and was originally purchased as a Turner, though unsigned. Art forensics expert Peter Paul Biro of Montreal, Canada has identified a fingerprint of Turner's in the paint, and our family is very interested in knowing whether your technology would be capable of giving us a second opinion and what would be involved in order to accomplish this. Would we have to bring the painting to Holland? Is your technology mobile? And what would be the fee to analyze a painting measuring approximately 3 ft. x 4 ft.?

Thank you for your time. We really look forward to seeing the documentary.

Kind regards. Anonymous

Postma: Dear Anonymous and Andrew,
Up to this point, our research has focused on Van Gogh, Rubens, and a limited number of other painters. The main two requirements for performing an analysis of the (dis)similarities of paintings are as follows:

  1. The analysis should be based on a sufficiently large set of paintings. Preferably, the set should contain works of the painter under consideration (e.g., Turner) and works of similar painters (same style and/or same period).

  2. For each painting we require an ektachrome, i.e., a large positive slide, created by a photographer under standard illumination conditions. Preferably, a color calibration card should be attached to the support of the painting. An example of such an ektachrome could be seen in the NOVA show when Laurens van der Maaten (one of my Ph.D. students) was holding up the ektachrome of Charlotte Casper's forgery for examination. Typically, ektachromes are created by the photography department of a (large) museum.

When we are provided with these materials, we will digitally scan the ektachromes and perform an analysis at no cost. Our benefit is that we acquire novel data and obtain experience with new paintings and painters. In return, you will receive a concise report on the results of the analysis. We will not publish or present the results obtained with your materials, or distribute the data beyond our group without your written permission.

You can send your material to the following new address (per Sept. 1, 2008):
Eric Postma
TiCC, Faculty of Humanities, University of Tilburg
PO Box 90153, 5000 LE Tilburg
The Netherlands

Q: One question that has baffled me is: Do experts who analyze the authenticity of art use the wood in the frame of the canvas to acquire results, or do they rely on the canvas and paint for the most part? Marc A.P. Vega II, Scottsdale, Arizona

Postma: Dear Marc,
In their analysis of paintings, art historians rely on many sources of information. They indeed often use analysis of the wood of the frame for dating (dendrochronology) and analysis of the canvas to determine the roll from which it came. They also analyze the owners, sellers, and buyers of a painting throughout its history (the so-called provenance), the microscopic properties of the painted surface, the chemical properties of the pigments, the visual properties of infrared and x-ray recordings of the painting, and so forth. Our algorithms provide an additional analysis tool, the results of which should be combined and integrated with those obtained from the other tools.

Q: Since Jackson Pollock did not use a brush to create his drip paintings, would your technique be able to tell a fake from an original? Teri Horton, Newport Beach, California

Postma: Dear Teri,
A very good question! Although we have not (yet) applied our techniques to Pollock's paintings, I am convinced that our algorithms could help to tell a fake from an original for two reasons. The first reason is that our algorithms do not rely on the explicit recognition of brushstrokes. Instead, they measure the visual properties of the brushstrokes (and their surrounding areas). In the case of Pollock, our algorithm would measure the visual properties of the patterns of "drips," which is so characteristic for his work. The second reason is that my colleague, Professor Richard Taylor at the University of Oregon, performed a so-called fractal analysis of Pollock's paintings and succeeded in authenticating his paintings (interested readers are referred to his Scientific American paper at http://materialscience.uoregon.edu/taylor/art/scientificamerican.pdf). Within our consortium (which includes Ella Hendriks, Ingrid Daubechies, James Wang, and Rick Johnson, who all appeared in the show), we are about to perform some analysis of Pollock's style in close collaboration with the MoMA.

Q: Can the same approach you use for paintings work for ancient items like statuary, painted ceramics, Egyptian wooden paintings, etc.?

Thank you. Anonymous

Postma: Dear Anonymous,
We have not yet applied our techniques to ancient 3D objects, but would love to do that. Provided that we have access to high-quality digital reproductions (e.g. ektachromes in different views), it wouldn't be too difficult to give it a try.

Q: Dear Mr. Postma,
In paint application where mediums are used for the refraction of light on the work, can your analysis detect if aging of the paint/medium causes the amount of refraction to change?

Thank you in advance, Albert

Postma: Dear Albert,
Provided that we have a sufficiently large number of (digitized) samples of aging paint/medium and their refection properties, our algorithms can learn to model the age-related changes in refraction. Of course, the main challenge is to obtain the examples. Since we cannot travel backwards in time, techniques to artificially age samples of paint have to be used.

Q: Do experts use paint chips to analyze the base paint to see if it's from a specific era? Ron, Noblesville High School, Noblesville, Indiana

Postma: Dear Ron,
As far as I know, experts do use color cards and paint chips for matching the type and age of paints, but I have to add that I am not an art expert.

Q: Hello. There is an inscription which is evidently on the reverse side of a very old painting which has been rebacked in the past as a conservation method. The inscription appears on the x-ray of the painting as a white area in which letters are detectable. This inscription is not visible on the painting itself, so I must presume that it is on the back side of the original canvas, sandwiched between two layers now. If the backing is removed, the inscription could be damaged, rendering it unreadable. Could your methods see through the canvas to help clarify the message of this lettering? I can send you a picture of what I have captured using my own methods. Thanks. Lynn Hildebrandt, Navasota, Texas

Postma: Dear Lynn,
Thank you for your email. Although our algorithms cannot "see through the canvas," we do have techniques that may enhance the visibility of the inscription. The success depends on many factors, but we can always try. Please send your picture to eric.postma@gmail.com and we will let you know what we can do.

Related Links

  • Art Authentication

    See if clever computer algorithms can distinguish a master forgery from a masterpiece.

  • Profile: Hany Farid

    This self-proclaimed "accidental scientist" is a digital detective inventing new ways to tell if photos have been faked.

  • Hany Farid: Expert Q&A

    Father of digital forensics and Dartmouth professor Hany Farid answers questions on photo fakeries and more.

  • Profile: Luis von Ahn

    A computer scientist finds novel ways to stop spammers and harness the brainpower of millions of people.